{
  "paper_id": "9973d524f2e90774d224356a51a1c40c",
  "paper_title": "DROW: Real-Time Deep Learning based Wheelchair Detection in 2D Range Data",
  "paper_authors": [
    "Lucas Beyer",
    "Alexander Hermans",
    "Bastian Leibe"
  ],
  "publication_date": "",
  "publication_venue": "IEEE Robotics and Automation Letters",
  "abstract": "We introduce the DROW detector, a deep learning based object detector operating on 2D range data. Laser scanners are lighting invariant, provide accurate 2D range data, and typically cover a large field of view, making them interesting sensors for robotics applications. So far, research on detection in laser 2D range data has been dominated by hand-crafted features and boosted classifiers, potentially losing performance due to suboptimal design choices. We propose a Convolutional Neural Network (CNN) based detector for this task. We show how to effectively apply CNNs for detection in 2D range data, and propose a depth preprocessing step and a voting scheme that significantly improves CNN performance. We demonstrate our approach on wheelchairs and walkers, obtaining state of the art detection results. Apart from the training data, none of our design choices limits the detector to these two classes, though. We provide a ROS node for our detector and release our dataset containing 464k laser scans, out of which 24k were annotated.",
  "keywords": [
    "2D range data",
    "deep learning",
    "CNN",
    "wheelchair detection",
    "walker detection",
    "laser scanner",
    "voting scheme",
    "depth preprocessing"
  ],
  "paper_institutions": [],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:14.774443Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文提出DROW检测器，一种基于深度学习的2D距离数据目标检测方法。传统激光检测依赖手工特征和提升分类器，性能受限于设计选择。我们采用CNN，通过深度预处理（中心化、截断、重采样到固定真实世界尺寸的窗口）和投票机制，有效利用激光深度信息，大幅提升检测性能。在轮椅和助行器检测上获得最先进结果，并在老年护理中心录制的包含464k次扫描（其中24k带标注）的数据集上评估，运行速度约75帧/秒。该方法可推广到其他类别，并已提供ROS节点和代码。\n### 关键要点\n- DROW采用CNN对每个激光点的固定真实世界尺寸窗口进行分类和回归，无需手工特征。\n- 深度预处理包括中心化、截断和重采样，确保网络感受野覆盖恒定真实世界范围，缓解距离变化带来的表示差异。\n- 每个窗口同时输出类别概率和局部偏移(Δx, Δy)的回归，通过投票网格和非极大值抑制(NMS)整合为最终检测。\n- 数据集包含464,013次扫描，其中24,012次标注了轮椅和助行器中心，分成训练、验证和测试集，且测试集来自未见的区域。\n- 在自身测试集和公开的Reha测试集上，DROW的准确率和召回率均显著优于Gandalf检测器。\n- 与直接预测整个扫描的YOLO风格基线相比，DROW的投票机制对检测性能至关重要，朴素CNN基线效果很差。\n- 消融实验表明，预处理中的重采样步骤对性能影响最大，去除此步骤后助行器几乎无法检测。\n- 使用局部偏移回归比回归角度和距离效果更好，回归角度和距离导致性能大幅下降。\n- 相比多层感知机和回归森林，CNN具有更强的空间先验，性能优于它们，且模型大小仅1.1MB。\n- 窗口大小选择1.66m时整体性能最佳，但更大窗口对轮椅检测更有利，提示多尺度输入可能进一步提升性能。\n- 检测器的定位精度高，评估半径0.3m以上性能稳定，表明检测中心预测准确。\n- 方法在距离3-5m范围内性能最优，符合机器人导航和规划的关键需求，且不依赖背景模型或位置先验。"
}